Github user xuchuanyin commented on the issue: https://github.com/apache/carbondata/pull/2314 @kumarvishal09 yeah, you are right. I've communicated with @ndwangsen and a new strategy will be added. Besides, the intention is mainly for enhancing data loading performance. With this strategy, carbondata will generate bigger carbon files to avoid too many small files. We ignore the data locality on purpose to avoid the situation that tasks may always run part of the executors. In a word, this strategy targets to loading small amount of input data.
---